It’s well known that GenAI has some serious biases in it, particularly for image generation. I’m currently producing images to include in a talk I’m doing, and without some significant prompt pushing, you’re way more likely to produce images of fellas with red hair than you are a person of colour.
The hoops you need to jump through to get a person who is obviously above the age of 50 is pretty astonishing too. Although if you ask for an image of a CEO, you’ll pretty reliably get an older looking man in a nice suit as can be seen below.
Prompt for an image of a CEO presenting to a board and you’ll frequently get
something like this
Without steering, this amazingly added women to the boardroom, although they are all white. I also have a sneaking suspicion that the lady on the right with what looks like red hair is supposed to be blonde and the colour grading pulled it towards red.
There’s nothing new here that hasn’t been said a thousand times already but having to push the systems so hard to get some diversity means that lots of people won’t bother and will just use what’s served up to them which reinforces entrenched biases.
This said, I’m sure the redheads out there will be pleased for some extra representation in my talk.